We have performed imaging studies of language organization in normal controls and patients with epilepsy. Using positron emission tomography (PET), activation of cerebral blood flow (CBF) associated with word and object recognition, auditory comprehension, and phoneme, word, and sentence production are localized in the brain. Studies of these functions in normals form the basis for evaluation of the effect of seizure disorders on cognitive processes subserved by temporal lobe and other cerebral structures. In studies of memory, deactivation of CBF during retrieval reflects the effects of earlier encoding. The deactivated regions are those which are engaged in the initial processing of stimuli. During a visual design recognition task, we found deactivation of right primary visual cortex. During an auditory recognition task, there was a relative CBF decrease in bilateral superior and middle temporal regions. We also evaluated a classical learning paradigm using eyeblink conditioning. We found learning specific increases in CBF in regions previously implicated in animal studies, including cerebellum, basal ganglia, frontal cortex, and hippocampus. Data from subdural stimulation, PET, and magnetic resonance imaging (MRI) are integrated using digital image processing techniques. The combined stimulation and PET data allow us to study the relationship between activation and disruption of cognitive activity, and to form more accurate concepts of the organization of cerebral function. These studies will elucidate the function of regions such as the basal temporal language area, which are of clinical importance when surgery for uncontrolled seizures is planned. Digital signal processing techniques are used to confirm anatomic localization of functional mapping. Using surface fitting algorithms, PET, CT, MRI, and subdural electrode positions are aligned. In PET experiments, rest conditions are averaged and subtracted from activated conditions, in order to reveal regions of increased blood flow during task performance. We found a high concordance between PET-CBF and subdural stimulation mapping using a number of different functional tests. This result shows the practicality of noninvasive preoperative functional brain mapping, and also demonstrates the close correlation of disruption and activation studies. We have found significant involvement of the basal temporal language area, which may explain unexpected postoperative deficits, in PET activation studies. During subdural mapping, stimulation of the basal temporal region disrupted implicit memory priming.